From Bugzilla Helper:
User-Agent: Mozilla/5.0 (X11; U; Linux ia64; en-US; rv:1.7)
Description of problem:
The existing tulip driver (0.9.15-pre12) in 2.4.21-20.EL
kernel has two known bugs:
1) In the ifconfig down path, tulip_remove_one() calls
pci_free_consistent() before calling unregister_netdev().
Fix is move the unregister_netdev() a few lines up in the source.
Jeff Garzik original sent me this fix more than 6 monthes ago.
This fix should be in all RHEL releases by now.
2) tulip_stop_rxtx() doesn't wait for DMA to fully stop like the
function call name implies. Charlie Brett (HP) gets credit for
finding this last April. I submitted a patch to Jeff Garzik but
don't recall it getting accepted or what the outcome was.
Patch is still available from:
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. from system A, generate large packets
(e.g. pktgen, "ping -b -f -s 1492")
2. on system B (HP ia64):
ifconfig eth4 down
ifconfig eth4 up
eth4 is a tulip device.
Actual Results: HP ia64 machine will MCA in minutes if no seconds.
Expected Results: system should not MCA.
Patch #1 is already in RHEL3 U4...
Patch #2 has not been accepted upstream (yet) -- will have to
investigate before I can push it in RHEL3...
Patches for #2 posted internall and upstream on 11/17...
HP has asked that this be fixed in rc1 if possible -- it is NOT a blocker.
Arg ! Wrong Release. HP asks if this can be in RHEL3 U4. It's too
late. Making U5 blocker.
A fix for this problem has just been committed to the RHEL3 U5
patch pool this evening (in kernel version 2.4.21-27.3.EL).
An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on the solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.